Innovative Approaches to Scalable and Multi-reference Coupled Cluster Methods

Increased computational speed and improved numerical algorithms have made computational chemistry an important tool in the development of new chemical compounds and processes. In particular, single-reference coupled cluster (CC) methods, such as CCSD(T), provide an excellent compromise between speed and accuracy in applications involving molecules near their equilibrium geometries, capturing most of the relevant dynamical correlation effects. However, single-reference CC methods may be less reliable for systems far from equilibrium or in characterizing electronic states with significant multi-reference character. At present, no such equivalent method is generally available for modeling these kinds of systems. We propose to address this need by developing a hierarchy of three CC methodologies designed to treat systems with a significant multi-reference character and to capture non-dynamical correlation effects. To reach a large community and to ensure the proposed approach is applicable to a large class of systems, the algorithms will be developed in a scalable (multi-processor) fashion and tested within GAMESS, a freely available, well documented, and popular suite of electronic structure codes. A wide range of chemical problems will be modeled and evaluated against many proposed and existing CC methods. Several benchmark calculations will be performed. BENEFIT: The proposed work will provide a unique technical capability which has several immediate military and industrial applications, including the development and testing of energetic materials for propellants and explosives, insensitive munitions, fuels, gas generators, chem-bio defense, advanced nanomaterials, and small-molecule drug design. In Phase I, existing and new methods will be evaluated, and a basic approach will be designed, tested, and evaluated within the GAMESS suite of electronic structure codes. Higher level methods will also be demonstrated and evaluated with expert level codes. By the end of Phase I, an evaluation of the relative speed and accuracy of a large number of CC approaches targeting non-dynamical correlation effects will be made. By the end of Phase II, several candidate higher level CC approaches will be fully implemented and benchmarked in GAMESS.